anova

anova is the multi-group mean-comparison node. It asks whether at least one group mean differs from the others.

It is not a pairwise explanation tool. A significant F-test tells you the groups are not all behaving alike; it does not tell you which pair caused the result.

Test contract

Field	Required	Notes
inputs	yes	Exactly one tidy table.
valueColumn	yes	Continuous numeric outcome.
groupColumn	yes	Column containing two or more groups.
groupLabels	no	Optional display/order hint for the groups.

How to read the result

default includes group summaries, between/within degrees of freedom, the F statistic, p-value, and effect size.

Use groupLabels when group order or display names matter in a report.

Assumptions and follow-up

• Watch sample-size, shape, outlier, and variance-ratio warnings next to the result.
• Plan follow-up pairwise t_test nodes only for comparisons you can justify, not every possible pair by reflex.
• For a non-parametric multi-group alternative, use a Python/R node for Kruskal-Wallis or permutation testing.

Example

- id: a
  kind: anova
  inputs: [data]
  valueColumn: outcome
  groupColumn: arm
  groupLabels: [a, b, c]          # optional

Related

• t_test - planned pairwise mean comparisons